Reinforcement Bar Spacing Detection Based on YOLO-Pose
To address the issue of poor accuracy of traditional methods for measuring steel bar spacing in construction sites due to complex backgrounds and noise,a novel detection method utilizing the YOLO-Pose algorithm is proposed.This method employs deep learning techniques to identify key points of steel bar intersections,exhibiting higher robustness and adaptability compared to conventional approaches.Through comparison among different detection networks,the YOLOv8-Pose model dem-onstrates outstanding performance in steel bar intersection detection tasks,achieving an average preci-sion of key point detection(mAP-kp)of 99.3％and frames per second(FPS)rate of 77.Experimental results indicate that this method,coupled with pixel calibration and diameter detection,enables precise calculation of steel bar spacing with an average relative error of 2.6％and a maximum relative error of 8.9％,complying with the acceptance criteria of the GB50204-2015 code for quality inspection of con-crete structure construction engineering.
reinforcement bar spacing detectionYOLO-Posedeep learningkey point recogni-tionpixel calibratio
万方数据
维普
